BacNguyen

Kaggle Competitions Master

[BacNguyen @ngxbac216]

@LionAayan Are you selling naked calls without holding enough shares to cover them?

[BacNguyen @ngxbac216]

@ifexy911 HIMS might visit 55 and 50 soon. Not many call in next two weeks.

[matthew sigel, recovering CFA @matthew_sigel]

🚨 Circle Rejected Ripple Takeover Bid of $4-5B

[BacNguyen @ngxbac216]

@gulvarol @tianyoulu @eccvconf I received an email stating that: "You have just been assigned a set of papers and your reviews are due on **Friday, April 26, 2024.**". However, no papers are appearing on my CMT portal.

[BacNguyen @ngxbac216]

@HermesLux Hey, I just curious why did you select the tickers with low volatility like this?. I see the bid-ask spread is too large, but it seems you were filled at a very good premium. Any ideas?

[#CVPR2026 @CVPR]

An important update on the #CVPR2024 submission deadline from the conference organizing committee:

Our Program Chairs have voted to shift the CVPR 2024 submission deadline to November 17th (A one-week extension). The website will be updated shortly to reflect this change.

[Thanh Dat Truong @dattruong_uark]

Our work "DirecFormer: A Directed Attention in Transformer Approach to Robust Action Recognition" has been accepted in #CVPR2022
This work presents a simple but novel perspective of Transformer-based approach to understand the right order of sequence actions in action recognition

[BacNguyen @ngxbac216]

@mrk3nx :)) bố bỉm sữa có khác

[BacNguyen @ngxbac216]

@greentheonly 2D box leads to inaccurate 3D location (x, y, z) estimation IMO. Fusion module takes both radar and vision info. It could be wrong if either is not precise enough. Thus incorrect matching.

[BacNguyen @ngxbac216]

@greentheonly I am curious the bounding boxes look different from the previous versions (the video you posted before). Did they replace 3D with 2D box instead?

[BacNguyen @ngxbac216]

@MaximilianMand1 @greentheonly The perception module would detect, track and predict the motion of the object over time. This attribute might not contribute to the final decision control IMHO. or there is a reason behind the scenes and they won't fix this bug.

[green @greentheonly]

Ok, autopilot assist app activated!

the much coveted augmented reality view!

(perfect "stop line" detection I must say, only 31% probability of indoors though which is a bit of a bummer.)
Interestign that A/B nodes vision fps is different (both far from 36fps nominal though!)

[BacNguyen @ngxbac216]

@dr_hb_ai @kaggle congrats

[BacNguyen @ngxbac216]

@abhi1thakur @kaggle Did you join Kaggle Team? :D

[Andrey Lukyanenko @AndLukyane]

Sometimes you can get a medal in #kaggle competitions, but sometimes you fail. It is important to learn from mistakes (just like neural nets) and improve. So I wrote a blogpost about the things which I have learnt hard way: https://t.co/AXp1y9afoD

[Sergey Kolesnikov @Scitator]

Thanks to Open Source tools, @PyTorch, @GoogleColab and @catalyst_core you can help fight COVID-19 right from the web, fully remote 💪
https://t.co/apw1JwWtxM

[BacNguyen @ngxbac216]

@ZFPhalanx i used metric learning to solve seen/unseen grapheme too. it does help on the private. but, my solution is not solid enough to get high score.

[BacNguyen @ngxbac216]

@catalyst_core @CenterTrends It is wonderful to hear about this project. Since my research relates to medical image processing (including the brain), would love to follow your work

[Quoc Le @quocleix]

AdvProp: One weird trick to use adversarial examples to reduce overfitting.

Key idea is to use two BatchNorms, one for normal examples and another one for adversarial examples.

Significant gains on ImageNet and other test sets.